In partnership with
Hi, there!
Welcome to the 17th edition of Work in Beta.
In this edition, we show you how to spot the work in your week that should become a Claude Skill - two lenses to identify it, three types to categorize it, and how to tell Claude what you want so it builds it for you.
Also, if you are looking to build your individual or organisational system with AI, scroll down to the bottom of the newsletter to know more and connect with us.
We're planning a Claude Code workshop soon. Hands-on, practical, built for people who want to actually work differently with AI, not just watch someone else do it. Drop your details here to get on the waitlist.
Let's dive in!
IF YOU ONLY HAVE 2 MINUTES
Image Credits: ChatGPT / Work in Beta
THE ‘HOW TO’ PLAYBOOK
You're Still Re-Explaining Your Own Workflow to AI
You have been using Claude for a while. It's helpful. But every session starts the same way - you explain the task, paste the same instructions, describe the format you want, correct the same things you corrected last time.
It works. But it's work about work.
A Skill removes that layer. (We covered what they are here) You define the procedure once, the structure, the format, the rules, and Claude applies it automatically. Every time. No re-explaining. No re-correcting. No re-pasting.
The result: a chunk of your week that used to take effort just happens. Consistently, correctly, by default. Not because AI got smarter. Because you told it exactly how you work, once.
The surprising part? Building a skill takes 15-30 minutes. Identifying what should become one takes even less. Two questions. That's all.
Two Ways to Spot Skill-Worthy Work
Not every task needs a skill. But more of your work qualifies than you might think. Two lenses help you spot it.
Lens 1: Does the structure stay the same even when the content changes?
Weekly reports. Client onboarding emails. Meeting summaries. Campaign briefs. Expense categorisations. The skeleton is identical every time - same sections, same sequence, same format. Only the content filling those sections changes.
That structure is what a skill captures. The content is still yours.
Lens 2: Does this take your time without needing your brain?
Reformatting data into a table. Applying a style checklist to a document. Following a 12-step review template. Enforcing brand guidelines across content. This work isn't creative. It isn't strategic. It just has to happen correctly every time and it eats hours without using any of your actual expertise.
If a task passes either lens, it's a candidate for building a skill. If it passes both, that's where the biggest time savings are.
Three Types of Skill-Worthy Work
Once you've spotted a candidate, it usually falls into one of three categories. This matters because it shapes how you describe the skill to Claude.
Type 1: Consistent Output Creation
Same structure, different content every time. The skill holds the template; you fill in the specifics.
Marketing: Weekly campaign performance summary - same sections, same metrics, different numbers each week
HR: Offer letters - same structure, different candidate details each time
Finance: Monthly expense categorisation - same categories, different transactions
Operations: Project status updates - same format, different progress data
Type 2: Multi-Step Workflows
A sequence that runs the same way every time. The value isn't in any single step, it's in the sequence being consistent.
Content: Blog production pipeline - research → outline → draft → edit → SEO check
Operations: Vendor evaluation - collect proposals → score against criteria → produce comparison → recommend
HR: Candidate screening - review applications → score against rubric → produce shortlist
Finance: Monthly close - gather data → reconcile → produce report → flag exceptions
Type 3: Embedded Expertise
Domain knowledge or quality standards that must be applied consistently. The expertise is the skill, not a format, not a sequence.
Marketing: Brand voice enforcement - every piece of content checked against your brand guide
Finance/Legal: Compliance pre-checks on contracts before human review
Operations: Deliverable quality criteria - every output scored against the same bar
HR: Performance review frameworks - consistent evaluation standards across the team
Here's a quick diagnostic. If you're always correcting format, you've got a Type 1. If you're correcting sequence, it's Type 2. If you're correcting standards, it's Type 3.
How to Tell Claude What You Want
You've identified a skill-worthy task. Now you need to communicate it to Claude so it can build the skill for you. Two paths. Both work.
Path A: The Conversation-First Method
This is the one we recommend and it's what Anthropic recommends too.
Do the task with Claude in a normal conversation. Correct the output until it's exactly right. Adjust the structure. Fix the format. Get the tone where you want it. Then tell Claude: "Turn what we just did into a skill."
Claude's skill creator - available in the plugin directory - packages the workflow into a skill for you. You didn't write a single instruction. You just did the work, and Claude extracted the pattern.
One blogger did this by pulling their last 7 articles and asking Claude to summarize the workflow they naturally followed - messy notes → draft → images → SEO polish. Result: 2 working skills extracted from their own observed patterns.
Path B: Describe It in Four Parts
If you already know your workflow clearly, describe it directly:
What goes in - the input (raw data, meeting notes, client brief)
What happens to it - the transformation (analyse, restructure, score, compare)
What comes out - the output format (report with 4 sections, summary email, scored rubric)
What to avoid - constraints (no jargon, max 500 words, never include client-internal data)
Then tell Claude: "Create a skill from this description." It builds the skill for you.
The Mistakes We See People Make
Mistake 1: Trying to build the perfect skill on day one. V1 is supposed to be rough. Use it 3-5 times. Every correction you make is feedback, that's your v2. One practitioner reported 3-4 rounds before smooth operation. That's normal. Skills are living documents. They get better the more you use them.
Mistake 2: Waiting until you're "ready." You can build and test a first skill in 15-30 minutes. The barrier isn't technical, it's starting. The best first skill is the one you need this week, not the one you've been planning for a month. Just get in the habit. The first one teaches you more than any guide will.
Mistake 3: Not reading what the skill creator gives you. If you use the skill creator to build a skill, read every single line of what it produces. Edit it. Challenge it. It's YOUR skill - it should reflect your standards, your judgment, your workflow. Not AI's generic interpretation of what you described. The skill creator gives you a draft. You make it yours.
Do This Week
Tomorrow: Look at your calendar from the past two weeks. Mark every task where the structure was the same both times, even if the content was different. That's your candidate list.
This week: Pick the most frequent one. Open Claude, do the task together in a normal conversation, and when the output matches what you want, say: "Turn this into a skill."
Next week: Use it 3 times. Note what still needs correcting. That's your v2.
Final Thought
A skill isn't a feature you configure. It's your own work, written down clearly enough that someone else could follow it. That "someone else" happens to be AI.
The people getting the most from Claude looked at their own week and asked one question: which parts of this should just happen?
WORK WITH US
The Other 95%
Knowing how to prompt well is roughly 5% of what it means to actually work with AI. The other 95% - context architecture, workflow design, thinking behaviors, tool orchestration - is where your workday actually changes. Not "I got a better first draft." More like "I rebuilt how my team runs weekly reporting, and now it takes 12 minutes instead of 4 hours."
That's what we work on with professionals and teams through Work in Beta.
For individuals: We help you build the skills, context files, and workflows that turn AI from a chatbot into a system that actually runs your work.
For organizations: We help teams design the AI operating systems - skills, workflows, context architecture - that make adoption stick beyond "everyone has a login."
If you're curious what the other 95% looks like, reach out to us here.